DXGI error fix (SOLVED??)

[pestridge77]

Phew... okay...about 7 days of testing P3D done. the results are in for V5.3

I tested two versions of P3D combined with various different drivers of Nvidia ALL on default settings of the sim with default V5, the only addons was FLAi, AI Reborn Pro and Chaseplane ...and the following settings in Nvidia:
Using a profile for p3D I always turned Vsync OFF in the control panel and set MAX FRAME RATE to match my monitors refresh rate of 60Hz. I used the ADAPTIVE setting for the power management. I set Texture Filtering Quality to HIGH. I selected the OpenGL Rendering to my GPU and changed the Phsyx setting to the GPU for all the tests.

So the versions of P3Dv5.3 ive been testing has been the original version 5.3.12.28072 and the latest version 5.3.17.28160 (all academic) with matching SDKs

I have not tested with version 5.3.15.28131 for the simple fact of the error on Dynamic Lighting set to ON all the time by default.

I tried various drivers dating back to 460.79 all the way up to 472.12
1. 460.79
2. 460.89
3. 461.40
4. 461.72
5. 466.27
6. 471.96 this is my favourite one.
7. 472.12
I havent tried the latest drivers as I am confident I would get errors anyway with them as LM would not have used those drivers when making the various versions of V5.3
I have tried to use the drivers that the LM testing group would have used or been on (using dates to match them of when they would have been building/testing) on the various versions.
... below is the process/environment I used to install and test within.

Windows 10 OS Build 19044.1503 Version 21H2
PC SPECS :
CPU _ i7700K running at 5.0Ghz with usage around 70-90%
GPU_ 1080Ti with stock speed and no overclocking applied running at around 65-90% usage
RAM_ one stick of 32G ( I know)
Two Monitors 1) 4K Samsung Curved 50inch with resolution set to 2K @60fps
2) 720P Monitor 27 inch with resolution set at 1080P @60fps
Temps for CPU are under 80c (60-75c) and temps for GPU are around 60-72c

I did three sets of tests, using both versions of P3D and all the 7 drivers with three major parameters that were changed across all the tests.

I would fly the F22 Raptor in VHHH at 14:00 with todays date. Winter selected and circle around the airport for 5 minutes (when it wasnt DXGI) and then fly off for 20 minutes and monitor the CPU temps, fps, CPU usage, GPU usage and temps and how much memory was being loaded in both the cache and physical.

1st Testing round (2 versions of V5.3 and about 7 different drivers =14 separate tests) I DID set ALL the drives to have allow EVERYONE all access.
2nd Testing round (2 versions of V5.3 and about 7 different drivers = 14 separate tests) I DID NOT set any permissions over any of my drives, I left them at default.
3rd Testing round (4 driver versions using both versions of V5.3 = 6 separate tests) I created a RAM Cache bat file that cleared the memory cache every ten minutes.
P3D is always installed in the default location it wants to install into.

As a usual set up I did the following so there would be no interference from the operating system:
I completely DISABLE Windows Defender/Security Suite and have no anti virus/live scanning running.
I DISABLE about 5 processes associated with Windows Update so there is no chance of the environment changing during the testing.
I also check the system with the cmd prompt sfc /scannow before installing each version of p3d and never was there any issues there during any of the testing Ive been doing over the last few days. I also cleaned my registry between tests each time and did a virus scan when changing the parameters of the permissions on the hard drives.
I also use (which I have always used since I created it back in V5.0) the Tdr delay regfix to increase or minimize the chances of DXGI when Windows detects spikes during all tests.

When re-installing my GPU driver, I would completely Uninstall the driver using DDU in safe mode with no internet connection. Then restart the PC and install the changed driver. Turn on Windows Defender/Windows Update and then download the Control Panel from the Windows Store. Restart and configure the GPU Control Panel with the settings above.


////////RESULTS:

So in total with the 7 different drivers I checked on, the two versions of P3Dv5.3 and in three testing environments.... the results were very interesting.

SEE BELOW FOR ANOMALY FINDINGS!! (different drivers gave me different headaches) Ray, the driver you mentioned was the best performing of them all.

1st ROUND OF Testing done in the environment of setting permissions in Windows on ALL drives to Everyone only I found the following errors:
DXGI errors more with 5.3.17.28160 on certain drivers although they still occurred once with 5.3.12.28072
Missing textures with both versions on many drivers.
missing flight deck textures until I reload the scenery or inflight more on 5.3.17.28160 if I try and set Unlimited fps in the sim.
very slow loading textures on 5.3.17.28160
content errors reporting missing textures for the F22 in 5.3.17.28160

2nd ROUND OF Testing done in the environment of having default Windows 10 permissions on all my drives only I found the following errors occurred:
Corrupted downloads from account on 5.3.17.28160 with random drivers
Corrupted unpacking of zip files for cabs scene2,8,12 every time on 5.3.17.28160 which I re-downloaded the individual ones from account.
Corrupted installation of installer which failed on later versions of drivers only tested with 5.3.17.28160
DXGI errors were on almost every driver installation of the latest 5.3.17.28160 and rare on the original 5.3.12.28072
In the sim I got missing textures on both versions, but more on 5.3.17.28160 with 4 out of the 7 drivers. Mainly the later ones.
Wrong textures being placed on wrong simobjects everywhere on 5.3.17.28160 with the last three drivers tested.
missing flight deck textures on both versions
content errors reporting missing textures for the F22 in 5.3.17.28160

3rd ROUND OF Testing done in the environment of using the RAM tweak and having permissions set to EVERYONE gave me the same results every time.
No DXGi errors when the using drivers 466.27 and 471.96 on any version of 5.3
content errors reporting missing textures for the F22 in 5.3.17.28160
Corrupted unpacking of zip files for cabs scene2,8,12 every time on 5.3.17.28160 which I re-downloaded the individual ones from account.




So, the results I got for me pin pointed a few things, which could just be particular to my set up between the two versions of P3D and the drivers used.... the DXGI error ONLY occurred when my VCRuntimes got corrupted and needed to be repaired once I had re-installed 5.3.17.28160. I had to update one of them that was absent as well, or when certain parameters were met.
the en_visual_cpp_redistributable_for_visual_studio_2012_update_4_x64_3161523 after which the DXGI error only returned when using the earlier drivers and CPU usage/temps was high, but less frequent. After fixing the VCRuntimes it was super hard to trigger a DGXI, even when trying classic DXGI triggering ways like ​maxing out the settings or running near high vram values.

I got random DXGI errors with earlier drivers more with the latest hotfix 5.3.17.28160 before the corrupted VCruntimes than with the original 5.3 on later driver versions, plus there were more texture issues, again, driver related issues. Which were the black textures everywhere and no flight deck textures/slow loading textures when slider was on Unlimited in later 5.3.

The VCRuntimes only corrupted once I changed the Permissions to default settings from the EVERYONE setting. Then I had corrupted VCRuntime errors when installing v5.3, either version.

*** ONLY received the DXGI error in both sims on ANY driver when my VCRuntimes was in need of repair of installation. *****
I have concluded that DXGI error can go towards for the following reasons, but not exclusive to any one reason, more of a mixture of reasons that creates a perfect storm for DXGI:
/The driver version used (makes a difference when combined with P3D version)
/If ANY file creates enough of a spike in vram to go beyond the hardware limits, no matter how long (rare and not that much of a factor)
/If the driver being used is not aligned with the version of p3D in use. (a major factor)
/If the VCruntimes are corrupted or missing for any reason (this seems to be the main trigger reason) ////////
/If the cache of the RAM get close to the limits of the actual RAM amount and isnt cleared, it increases the amount of DXGI errors that can occur.
/A possible corrupt driver install (when the Control Panel crashes when accessing the Program Settings tab) there is a work around.
​/When CPU usage is nearer 100% and the temps go into the 90s for longer duration. (which would then limit the GPU)


whereas the Tdr delay fix only provides very limited protection, it isnt a fix, more of a necessary precaution I would say using the older drivers.. There seemed to be two golden drivers that DXGI didnt occur on much when using the RAM tweak and setting permissions to EVERYONE.
When using the RAM tweak of clearing out the cache every ten minutes, I rarely got any DXGI error. the CPU remained around the 75% usage mark and the GPU temp never rose above 75c. All other times, the CPU temps went into the 90s and the GPU was in the high 80s.
The major factor is the vcruntimes.
I did notice when using the driver version 466.27 the little hand on the mouse in sim, was compressed/squashed than when using any other driver.


So my tests conclude to me that a DXGI can be triggered if you want too and can be avoided as well... DXGI happens when:

****CPU usage goes high in temp (into the high 80s/90s) for long durations, you will increase the chance of a DXGI as the GPU becomes limited in use.
****If the driver version you are using is far away from the v5.3 version of when it was released, you increase the chances of it happening even when CPU temps are between 60-75% in sim and GPU temps are 50c on the GPU
**** If there is a spike in vram, even with the Tdr Delay: this only happened on older drivers with the latest version of v5.3 not very often.
****VCRuntimes become corrupted or arent present... 100% guaranteed DXGI
**** If the sim doesnt close down properly and the process is still running when you load another process, more DXGIs can happen with older drivers and Chaseplane crashes.
****Increase in DXGI in recent 5.3.17.28160 when shadows are present on Vegetation for cast and receive




ONE THING STOOD OUT THO IN 5.3.17.28160 on my system......
*If in sim FPS rate is set to anything but 30fps (Vsync ON or OFF) the flight deck has black textures and other textures slowly load in. SET FPS TO 30 *IN SIM and everything is fine. (Vsync ON or OFF) on ANY DRIVER (again, may only be to my system and GPU/driver tho)
*When the mouse is on the screen, there is a dip in FPS in the sim. (could be the driver)
*Vsync in sim set to ON raises the usage of the CPU.
*Setting Unlimited to fps slider increases the usage of the CPU and over time raises the temps of the cores.
*Setting the Vsync in the GPU lowers the fps rate when Adaptive, ON or Adaptive half refresh rate is selected. (OFF with MAX Frame Rate set works
best)
*When starting on the runway the steering tries to veer to the right when the brakes are released and the steering has to be half left to compensate, maybe that tho is a feature put in.


So using this info, everyone should be able to avoid DXGI errors from occurring but hard to detect when the VCRuntimes get corrupted or need installing.

[JorgenSA]

The whole credibility of this long diatribe goes down the drain with these words:

"I havent tried the latest drivers as I am confident I would get errors anyway with them" - then the words after that in the rest of the sentence just serve as an excuse.

And BTW - VCruntime errors are avoided by:

1. Installing ALL the C++ Redistributables from Microsoft, and

2. Keeping them updated, using Windows Update, plus

3. Only running known compatible add-ons.

Edit PS: Testing with add-ons is also, scientifically speaking, a no-no, as these may introduce or contribute to the error(s). Testing for such an issue must be done with a virgin installation.

Jorgen

[pestridge77]

yes thats right Jorgen, updating the C++ Redistributables or running a repair on them is also helpful.. although I was able to get a DXGI after repairing the C++ Redistributables and before installing that one VCRuntime.. once I installed that, ALL the DXGI I was getting, went away...
These tests that I did, I had these results... someone else may come along and have different results...which is exactly what i found out, there are multiple ways to avoid these DXGI errors. its a mixture of things that make them happen... although one major cure that literally eliminated them.
Im sorry you didnt read further down the document although having recent up to date drivers has never served me well in p3dv5, Ive had more crashes than not when I did... of course its not advisable to not update the drivers, especially if anyone plays games as well, I am talking from a p3d point of view....
Plus, it kind of makes sense... if a version of p3d was compiled in a certain environment and tested using various drivers, how can it be compatible with driver versions not yet even made? I will tho test the latest Nvidia driver for my card, the 511.65, just to satisfy what you say, I will add my results as a comment on here. Surely the program becomes compatible with the driver when it comes out before it?? Not after. P3D only release updates about maybe once or twice a year, three if there are issues in a recent release, drivers get updated almost every three months, so how can P3D be made to work with future drivers, months down the line?

PS to your PS: You are correct with the addons, sadly, without having Chaseplane, the testing becomes more difficult and longer with the default camera system to move around..and I couldnt run the sim without having AI Reborn Pro in with Flai as errors came up. Everything else was default tho. I even kept Bathymetry ticked....Im sorry you dont appreciate the hard work that went in to help others .

[JorgenSA]

I would hold off on testing for a little bit - about 20 minutes from the time of writing - as Windows Update is supposed to come out at 19:00 or 7:00 PM my time, 10:00 AM US Eastern time.

Jorgen

[pestridge77]

Would that not make the test with the latest driver less valid against the others?

Edit : Updating now and will try the new driver

[pestridge77]

I would hold off on testing for a little bit - about 20 minutes from the time of writing - as Windows Update is supposed to come out at 19:00 or 7:00 PM my time, 10:00 AM US Eastern time.

Jorgen

okay @jorgen

I installed the new driver and updated Windows..

Not sure what to say.

The latest version of p3D is still plagued with the issues I had in previous drivers with higher temps in the CPU, whereas the first 5.3, I didnt get any of these issues. Slow texture loading as well as missing building textures in cities... alot more issues than the initial side gsx jetways issue of 5.3. For me, the initial 5.3 was very good. maybe the driver I was using was better for it.

The driver version 511 itself, seems okay, although I get alot more DXGI errors in the latest version of 5.3...which I think is more to do with the current 5.3 than the previous one, whereas I never saw one in the first initial of 5.3, I think I will roll back to 5.3 from the latest and simply await 5.4.
Increased CPU usage, hence higher temps and the DXGI increase, it aint worth it.. with the latest 5.3, the temps of the CPU hit the 90s and low 80s, all the time, not just on loading. In the initial 5.3, they never went above 80c.Ive tested the cooling and its working fine... I gave the PC a rest as well so it could recover the CPU from all the recent testing I did.
I just hope the original 5.3 works with this updated driver, otherwise I will have to roll that back as well.

I also notice the ADDONS tab in the drop down menu, is gone when I installed the latest driver?
Which I found very weird... how would that effect the sims menu>?

Is it a case of if there are no addons in the xml file, then the menu doesnt show?

EDIT: The ADDONS menu came back after repairing some Runtimes/C++redistributables